Clarify the description of mapping_provider

[matentzn]

          I think the description needs to be updated here. These URIs don't actually point to resources, so what are you supposed to do with this? This is more of a comment string since I'm not sure there's a good way to actually capture this information unless it's supposed to point to a metadata file or the source data itself

Originally posted by @cthoyt in #394 (comment)

[nichtich]

Required by gbv/jskos#108

[matentzn]

The current definition is:

mapping_provider:
    description: URL pointing to the source that provided the mapping, for example
      an ontology that already contains the mappings, or a database from which it was derived.
    range: uri
    instantiates: sssom:Propagatable
    annotations:
      propagated: true
    examples:
      - value: https://www.ohdsi.org/
        description: A URL pointing to the Observational Health Data Sciences and Informatics initiative.
      - value: https://monarchinitiative.org/
        description: A URL pointing to the Monarch Initiative Resource.

Assuming you are both after a glass clear rationale why URL vs URI/RDFS resource.

I personally wanted to keep this as open as possible. What I originally intended is of course using a resource id, like an infores CURIE (e.g. infores:go, ror:123), but for the groups I was working with it was too much of a barrier (like OHDSI) - they just simply would not fill in this field if I would ask for it to be a resources identifier.

I dont know what is best... Opinions?

[nichtich]

The examples don't match the description "for example an ontology that already contains the mappings, or a database from which it was derived." Both https://www.ohdsi.org/ and https://monarchinitiative.org/ are projects/groups/initiatives that happen to publish ontologies or databases. My confusion is also about to mapping_source / mapping_set_source and how it relates to mapping_provider.

Reading the examples, I'd think that a mapping_provider (some agent) is the publisher of a mapping_source (some dataset) but the definitions more look like mapping_provider (any kind of dataset) is a superset of mapping_source (only special case MappingSet) so both can have same values?

[matentzn]

Good point. The original use case for this field was to denote a resource, like "uniprot", OMOP vocabulary server, BARTOC etc, but it quickly devolved (probably due to the poor description) to be used for organisations. I think on average, this property was mostly used to denote the organisations providing the mappings.. Not sure what is best here, but maybe if we extend the definition towards including projects/groups/initiatives as well it would help?

[nichtich]

Not sure what is best here, but maybe if we extend the definition towards including projects/groups/initiatives as well it would help?

Yes, how about "URI/URL of the organization or service that provided the mappings, for example a publisher, repository or database from which they were derived"

In JSKOS this would broadly be:

SSSOM	JSKOS	RDF via JSKOS
maping_provider	publisher	dct:publisher: An entity responsible for making the resource available.
mapping_[set_]source	source	dct:source: A related resource from which the described resource is derived

The question whether a specific database is more a "source" or more a "provider"/"publisher) can only be answered for the individual case.

[matentzn]

Yes, how about "URI/URL of the organization or service that provided the mappings, for example a publisher, repository or database from which they were derived"

I like this. WIlling to make a PR?

https://github.com/mapping-commons/sssom/blob/master/src/sssom_schema/schema/sssom_schema.yaml